MXuer / Federated-learning-ASR

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Federated-learning-ASR

In this repository, you can find the code used to generate the results for our ITG 2021 publication. The espnet and kaldi recipes are in the corresponding subdirectories. They run independently from each other. We uploaded our code modifications and some parts of the underlying systems, but not the complete frameworks. These can be found in the respective repositories:

  • https://github.com/mravanelli/pytorch-kaldi
  • https://github.com/kaldi-asr/kaldi
  • https://github.com/espnet/espnet

Please make sure to provide the missing components in the correct folders, e.g. the complete pytorch-kaldi source code in the folder with the same name. The same needs to be done with the espnet sources.

Cite as

@inproccedings{yu-itg2021,
title={{Federated learning in ASR: Not as easy as you think}},
author={{Wentao Yu, Jan Freiwald, Sören Tewes, Fabien Huennemeyer, Dorothea Kolossa}},
journal = {ITG Conference on Speech Communication},
year=2021,
month = {October}
}

About

License:MIT License


Languages

Language:Python 41.6%Language:Shell 36.6%Language:Perl 21.8%